Climate change communication to the public is an important component of the “dual-carbon” policy system. The use of framing strategies in communication can affect public perception and behavior, thereby improving the effectiveness of communication. However, the existing research on the framing strategies in climate change communication is insufficient in the use of media and method to support innovation and optimization of strategies. This paper extracts subscripts of 234 climate change documentaries from the video media YouTube to construct a corpus of 1.3 million words, which is analyzed by an unsupervised machine learning approach-analysis of topic model networks-to identify strategies. We find three strategies in climate change communication videos: environmental threat frame, human threat frame, and controllable frame. The first two emphasize the threat of climate change to the environment and human beings, respectively. The latter emphasizes that the negative impacts of climate change can be controlled through technology and multi-actor participation, and is a strategy not being identified in the existing literature. This paper has important implications for promoting the practice of climate communication in China.